Facilitation and biodiversity–ecosystem function relationships in crop production systems and their role in sustainable farming

We review the need for increasing agricultural sustainability, how this can in part be delivered by positive biodiversity–ecosystem function (BEF) effects, the role within these of plant–plant facilitation, and how a better understanding of this role may help to deliver sustainable crop (particularly arable) production systems. Major challenges facing intensive arable production include overall declines in biodiversity, poor soil structure and health, nutrient and soil particle run‐off, high greenhouse gas emissions, and increasing costs of synthetic inputs including herbicides, pesticides and fertilisers. Biodiversity–ecosystem function effects have the potential to deliver win–wins for arable food production, whereby enhanced biodiversity is associated with ‘good outcomes’ for farming sustainability, albeit sometimes through negative BEF effects for some components of the system. Although it can be difficult to separate explicitly from niche differentiation, evidence indicates facilitation can be a key component of these BEF effects. Explicit recognition of facilitation's role brings benefits to developing sustainable crop systems. First, it allows us to link fundamental ecological studies on the evolution of facilitation to the selection of traits that can enhance functioning in crop mixtures. Second, it provides us with analytical frameworks which can be used to bring structure and testable hypotheses to data derived from multiple (often independent) crop trials. Before concrete guidance can be provided to the agricultural sector as to how facilitation might be enhanced in crop systems, challenges exist with respect to quantifying facilitation, understanding the traits that maximise facilitation and integrating these traits into breeding programmes, components of an approach we suggest could be termed ‘Functional Ecological Selection’. Synthesis. Ultimately, better integration between ecologists and crop scientists will be essential in harnessing the benefits of ecological knowledge for developing more sustainable agriculture. We need to focus on understanding the mechanistic basis of strong facilitative interactions in crop systems and using this information to select and breed for improved combinations of genotypes and species as part of the Functional Ecological Selection approach.


| INTRODUC TI ON
Facilitation is taken here to be interactions, either direct or indirect, between two or more neighbouring plants with a beneficial outcome for at least one of the neighbours (Bronstein, 2009;Brooker et al., 2008). Important consequences of facilitation are enhanced overall plant diversity (e.g. Choler et al., 2001;Butterfield et al., 2013), and enhanced ecosystem function, either directly or as a consequence of enhanced diversity (e.g. Boudreau, 2013;Li et al., 2007;Losapio et al., Submitted;Lozano et al., 2017). The simultaneous impact of facilitation on biodiversity and ecosystem functions indicates that facilitation may be playing a role within BEF relationships. Many studies have been conducted to investigate such relationships (see e.g. Soliveres et al., 2016;Tilman et al., 2012), and review papers, building on this large body of work, have looked for general principles. For example, Cardinale et al. (2012) concluded that biodiversity loss reduces ecological process efficiency and ecological stability through time, and that 'change accelerates as biodiversity loss increases'. O'Connor et al. (2016) found that an average BEF relationship from across 374 experiments 'characterized the vast majority of observations, was robust to differences in experimental design, and was independent of the range of species richness levels considered'. Furthermore, recent analyses suggest experimental studies undertaken on BEF effects can provide robust and more widely applicable information, despite in some cases being undertaken in 'unrealistic communities' (Jochum et al., 2020).
However, a lack of interdisciplinary collaboration between BEF scientists and crop scientists may be hindering progress in this area, and we have a poor understanding of the operation of positive diversity effects in intensive agricultural systems (Mason et al., 2017) and therefore how to enhance them. Our aim here is to explore the role of facilitation in BEF effects with particular reference to arable production systems, that is to say farming systems focussed on the production of crops. Arable systems are a vital component of human food production but are experiencing substantial challenges in terms of long-term sustainability and resilience to future environmental change. Critically, facilitative plant-plant interactions might be a key element in addressing these challenges (Brooker et al., , 2016.
After briefly reviewing the major challenges facing arable agriculture, we consider why BEF effects may be part of the solution, the potential role within these of facilitation, and how we might enhance desirable facilitation-based BEF effects to promote sustainable agricultural practice.

| THE NEED FOR IN CRE A S ING AG RICULTUR AL SUS TAINAB ILIT Y
The challenges for modern arable agriculture, particularly in regions and countries practising widespread 'intensive' agriculture, are considerable. The market demands that agriculture delivers affordable crop products of standardised quality to processors, food manufacturers, retailers and the public. To respond to this challenge, industrialisation of farming has become widespread, relying on mechanisation (reducing labour costs), agrochemicals (to increase yield and control pests and diseases), and within-field and withinfarm specialisation to increase efficiency and reduce costs. These changes have been associated with simplification of farming systems, which is evident in the widespread growing of a small range of crop species, typically in genetically identical stands of a single crop variety , and a reliance on only a very small amount of the plant biodiversity available. Of the approximately 50,000 edible species of plants, 150-200 are actually frequently consumed, and 3 provide 60% of the calories in the human diet (maize, rice and wheat; IPES-Food, 2016). These long-term changes in agricultural practice-in particular since the Second World War (Robinson & Sutherland, 2002)-have accelerated greatly in recent decades thanks to advances in crop breeding, the production of synthetic chemicals and mechanical and digital technologies.
The consequences of these moves towards industrialised agriculture, which in many cases have had to be adopted by farmers to remain competitive, but which at the same time may be difficult to reverse because of associated system simplification, are widereaching and multi-faceted. As summarised in Table 1, they include declines in farmland biodiversity because of, for example, improved mechanised seed cleaning, increased habitat disturbance and reduced habitat complexity, and the negative impacts of herbicides and pesticides on the wider environment, increased pollution (e.g. nitrate and phosphate run-off in water courses) and substantial contributions to GHG emissions (Critchley et al., 2006;Robinson & Sutherland, 2002;Storkey et al., 2012).
In addition, such farming practices are storing up problems for the future. With respect to climate change, as well as general trends basis of strong facilitative interactions in crop systems and using this information to select and breed for improved combinations of genotypes and species as part of the Functional Ecological Selection approach.

K E Y W O R D S
biodiversity-ecosystem function relationships, crop breeding, functional ecological selection, pest and disease resistance, plant-plant facilitation, review, soil nutrients, sustainable crop production towards warmer climates and changes in patterns of rainfall (with growing seasons becoming wetter or drier, depending on location), climate change scenarios indicate increasingly variable and extreme weather (e.g. UKCP, 2018). Beyond climate change, there are also general concerns about the future fragility of global food supply chains, a risk highlighted most recently during the Covid-19 crisis (Laborde et al., 2020). Many affluent countries are now dependent on imported foodstuffs to supply a considerable proportion of the TA B L E 1 Examples of the challenges currently facing modern agriculture that might be addressed by the facilitative effects found in crop mixtures (including examples from cultivar and species mixtures), details of the underlying mechanisms of these effects (with associated references) and the benefits they bring to crop production. Effects are subdivided into direct facilitative effects (impact of plant A on plant B mediated by changes in the abiotic environment) and indirect facilitative effects (mediated by the intermediary action of other organisms, including soil organisms, invertebrates, pathogens or other plants)

Challenges for sustainable agriculture
Facilitative effect Underlying mechanism

Benefits to crop production
Direct facilitative effects Efficient use of fertiliser, reducing cost, GHG emissions and potential risk of fertiliser run-off Enhanced nutrient supply Direct transfer of N from N-fixing legumes to non-legumes White et al., 2013) Increased availability of P, either through soil acidification by legumes (Cu et al., 2005), or the release of phosphate mobilising compounds Li et al., 2016) Increased productivity by increasing total available resource pool Efficient use of irrigation water to reduce GHG emissions and negative impacts on local water supplies Enhanced water supply Hydraulic night-time uplift of water by deeprooted plants and subsequent provision to neighbouring shallow-rooted plants driven by a water potential gradient (Caldwell et al., 1998;Izumi et al., 2018;Pang et al., 2013;Prieto et al., 2012) Increased productivity by increasing total available resource pool

Reduced water demand
Dense and complex canopies lead to reduced windspeeds, lower boundary layer conductance, local humidification of air surrounding leaves, and reduced leaf water demand and loss (Meinzer, 1993;Vincent et al., 2017;Yin et al., 2020) Increased productivity by increasing total available resource pool

Indirect facilitative effects
Reduced use of fertiliser and pathogen control agents; reducing net GHG emissions through increased soil C and reduced machinery passes; reducing run-off and loss of soil and nutrients Enhanced diversity and function of soil organisms Increasing plant diversity increases complexity of soil structures and substrates, positively impacting on diversity and function of soil organisms and processes, including C storage, nutrient release, and soil drainage and aeration (Burrows & Pfleger, 2002;De Deyn et al., 2008Solanki et al., 2019;Song et al., 2007) Improved soil condition and functions supporting crop growth; Increased productivity by increasing total available resource pool Reduced use of crop protection products and lower GHG emissions from reduced use of machinery

Control of pests and diseases
Reduced availability of susceptible hosts, less efficient dispersal, and altered microclimate (Boudreau, 2013;Newton & Guy, 2009)  population's diet, which itself then feeds back to enhance the scale and rate of climate change (Pradhan et al., 2020).
For all these reasons, as well as an increasing focus on whether our food production systems are targeting the right goals (Benton & Bailey, 2019), there is now considerable effort to develop genuinely sustainable approaches to agriculture. These would enable yield gains-or at least maintenance of yields-while reducing negative environmental impacts and providing resilience for the future, both in terms of the impact of climate change and other future shocks to global supply chains. Critically, as can be seen from the multiple demands on future agriculture, the optimum situation would be to identify and develop win-wins, that is farming practices that enable us to address multiple needs simultaneously.
Such approaches are in some ways recreating the multiple benefits that arise from traditional farming practices such as the wellknown 'three-sisters' polyculture of maize, bean and squash (see e.g. Zhang et al., 2014). The benefits of polyculture approaches are demonstrated by their widespread use: as Vandermeer (1989) points out 'If so many traditional agriculturalists do it, there must be some advantage to it'. A more modern approach to delivering and enhancing the benefits of multi-species cropping systems is the 'doublehigh' approach being developed in Chinese agriculture, which focusses on achieving both high crop productivity and high resource use efficiency through optimal crop system design and management, high nutrient use efficiency, improving soil quality and minimising the ecological footprint (see e.g. Shen et al., 2013). Importantly, both these traditional and modern approaches are in contrast to historic trends which favoured yield over environmental sustainability and system resilience. Ideally, we would move to a state of regenerative agriculture where, rather than halting further decline, agriculture plays an active role in improving the health of the system, for example by building reserves of soil carbon and biodiversity, thereby promoting long-term sustainability.

| THE ROLE OF B EF EFFEC TS IN DELIVERING AG RICULTUR AL SUS TAINAB ILIT Y
Critically, some of the general features of BEF relationships (Cardinale et al., 2012;O'Connor et al., 2016;Tilman et al., 2014) indicate the potential for win-wins, whereby enhanced biodiversity can generate benefits in farming systems in terms of enhanced functions (e.g. Minns et al., 2001;Snapp et al., 2010;Vandermeer et al., 2002). A key feature is the generally positive shape of BEF relationships. When combined with the commonly asymptotic shape of these relationships, this means that increasing biodiversity in those systems which are biodiversity poor is most likely to have positive impacts on function . Highly depauperate systems would include, for example, monoculture crops where both crop and wider diversity (e.g. weeds, invertebrates including pollinators and soil organisms) have been reduced by farming practices.
There is a considerable evidence of the beneficial effects for system sustainability of enhancing biodiversity in farming systems (Gurr et al., 2016). Farmland biodiversity can be enhanced at a range of scales, including the landscape, habitat, within-field and microbial scales (Newton et al., 2011). Integrated Pest Management (IPM), for example, is enabled in part through enhancement of biodiversity at these larger scales. This can include the reservoirs of biodiversity in semi-natural and managed environments adjacent to the crops as well as in hedges and field margins, and in specific measures such as beetle banks and floral strips managed within the immediate crop production environment (Birch et al., 2011). Although some of these effects might be considered facilitative (albeit long-distance and indirect), here we focus on the crop and associated within-field management as the most important primary unit of the arable production system.
Of particular interest here are studies of crop mixtures (Letourneau et al., 2011). Crop mixtures are simply the growing of two or more crops together and can include cover crops as well as cash crops. Cover crops are planted for benefits that do not arise from a final harvest of the crop itself, but arise instead from having some kind of 'cover' on the land, including preventing soil erosion and improving soil health, soil nutrient status and drainage (Bergtold et al., 2017;Snapp et al., 2005). benefits (Brooker et al., 2017) and-specific to IPM-mitigating selection for fungicide resistance by enhancing disruptive selection in the pathogen population, thereby enhancing crop protection and yield (Kristoffersen et al., 2020).
With respect to cover and cash crops, facilitative interactions can take place within cover crops as well as with the final cash crop, and the agronomy of the cash crop may utilise the protection of the cover crop, for example to enhance establishment: the legacy of the cover crop effectively facilitates the cash crop, though some of its mechanisms may be spatially or temporally separated from the cash crop (see e.g. Barel et al., 2019;Snapp et al., 2005) (Kirkegaard et al., 2008;Robson et al., 2002).
The facilitative benefit however is realised in the succeeding cereal crop harvested in the following growing season.
While we can find these beneficial effects, we should not assume that they are without any associated problems or costs. We know that-as for general BEF effects-they are often context and scale-dependent (Li et al., 2020), and highly variable in magnitude Vandermeer, 1989

| DEFINING THE ROLE OF FACILITATI ON A S PART OF B EF EFFEC TS IN AG RICULTUR AL SYS TEMS
Facilitation clearly plays a role in delivering the benefits arising from enhanced crop diversity, including in crop mixtures (Brooker et al., , 2016Li et al., 2014). Facilitative processes in arable systems can be direct and indirect and occur above-ground and below-ground. So, whether the BEF relationship associated with these 'good outcomes' is positive or negative depends on the metric used to assess function. Notably, many of the metrics used to assess crop system performance (net productivity, resource use efficiency, C storage) are used to assess function in studies of other ecosystems (see e.g. Tilman et al., 2014) and, as in these other systems, they tend to show positive BEF relationships in crop systems, certainly when starting from a low level of initial species diversity.
In some of the cases shown in Table 1, there is clearly a beneficiary and a benefactor; for example legumes provide fixed nitrogen (N), deep rooted species deliver hydraulic uplift, or the presence of a pathogen-resilient cultivar or species enables pest and pathogen dilution, barrier and induced resistance effects. In other cases, the effect is non-specific, with a general increase in biodiversity having the potential to enhance functions, for example soil C storage, structure and function (He et al., 2009). Because some crop systems are highly depauperate in biodiversity, even if some effects are relatively nonspecific the response to increased biodiversity may be substantial. But precisely what type of BEF effects are they? An overall positive BEF effect occurs when addition of a unit of biodiversity (genotype, functional type, species and habitat) enhances the function of the system . Within the BEF literature (see, e.g. Loreau & Hector, 2001), several types of underlying mechanistic effects are also defined. Sampling effects result from the increased probability that a more species-rich system will include at least one species with comparatively extreme trait values and hence the potential to impact ecosystem function. For this potential to be realised, selection or complementarity effects then need to occur.
Selection effects operate when the influential species runs to dominance in the community; they can be both positive and negative, for example when a comparatively productive or unproductive species runs to dominance, respectively. Complementarity effects result from either a more complete use of available niche space through niche differentiation (with gaps in niche space being filled as species, and hence functional diversity, increases) or from facilitation, noting the difficulty of distinguishing between niche differentiation and facilitation in practice (Loreau & Hector, 2001). Barry et al. (2019) also point out that many studies conflate the higher-level category of complementarity effects-which can have several underlying mechanistic causes, including direct and indirect facilitation-with the benefits arising from niche differentiation (sometimes called niche complementarity). Importantly, with selection effects the overall productivity of the community can be enhanced if a particularly productive component runs to dominance but cannot exceed the productivity of the best performing component in monoculture. In the terminology of crop production, selection effects cannot result in transgressive over-yielding . In contrast, complementarity effects, including niche differentiation and facilitation, can lead to transgressive over-yielding (Loreau, 2004 Recent studies are showing that in crop systems complementarity effects are of a similar scale to selection effects (Engbersen et al., Submitted), and what is perhaps most important is to recognise that facilitation is playing a very substantial (albeit sometimes complex and interactive) role within overall complementarity effects. Therefore, we must think about facilitation explicitly when considering how to manipulate and manage BEF effects to help deliver sustainable or regenerative agriculture and achieve a 'good outcome' for the system.

| INTEG R ATING FACILITATI ON -DRIVEN B EF EFFEC TS INTO SUS TAINAB LE AND REG ENER ATIVE AG RICULTUR AL THINKING A ND PR AC TI CE
A potentially hidden consequence of bundling together facilitation and niche differentiation within complementarity effects, although driven by the practical challenge of experimentally or mathematically isolating them (Loreau & Hector, 2001), is that it may have led to oversight of the role of facilitation in productive systems such as arable crops, leading to missed opportunities for the application of new analytical or conceptual approaches. Recognising the important role of facilitation in BEF effects that promote sustainable agriculture helps us in two key ways. In line with this, Chen et al. (Submitted) showed that current modern cultivars (bred for self-competition in monocultures) show a reduced reproductive effort in mixed systems compared to monocultures.

| Linking fundamental ecological understanding of facilitation to breeding for sustainability
However, there is no indication of similar effects in natural grassland species (Roscher & Schumacher, 2016).
The importance of evolving in a mixture for enhancing the beneficial interactions that occur is of huge relevance to designing and breeding for sustainable crop systems which involve mixtures. There is already concern that the process of crop selection and breeding, with a focus on yield quantity, quality and uniformity, has led to the loss of many traits that might help crop plants deliver multiple ecosystem services, or deal with the increasingly severe and variable environmental conditions expected under climate change, or with the reductions in inputs needed as we move towards more sustainable agriculture (Chacón-Labella et al., 2019;Milla et al., 2017).
For example, there is evidence that traits beneficial for organic or non-inversion tillage can be lost in modern breeding  Importantly, while needing to explore this issue in more detail, we already know of some traits that underpin facilitation, as indicated in Table 1. We are also getting increasingly detailed knowledge of where in the available crop germplasm we might still find the genes needed to alter these traits and, therefore, promote facilitation. A good example of this is recent work on traditional landraces. For example, Cope et al. (2020) and George et al. (2014) have undertaken studies of traditional landrace (known as bere) barley cultivars. These studies have demonstrated how bere cultivars are able to increase the availability of a wide range of nutrients on nutrient-limited soil. Such studies help us both to identify breeding targets and to understand how these effects are in part mediated through the close relationship between the barley plant and soil organisms. The advantage of using ancient landraces for plant genetic improvement has also been demonstrated in studies investigating the molecular basis of adaptation to high soil boron in wheat landraces (Pallotta et al., 2014), aluminium tolerance in barley landraces (Fujii et al., 2012) and phosphorus efficiency in landraces of rice (Gamuyao et al., 2012). However, how such landraces interact with other species in intercrops is yet to be studied, although impacts of landraces of wheat on suppression of weeds through shading have been proposed (Murphy et al., 2008). Screening landrace collections and heritage varieties of crops, both alone and together in intercrops, may be a good first step to identifying and understanding some of the relevant functional diversity.
A focus on below-ground traits may also be particularly important. We know that root traits can be critical in enabling crop cultivars to cope with the increased stresses associated with a less intensively

| New analytical frameworks for facilitation in crop systems
The second major benefit of recognising the role of facilitation in BEF effects underpinning sustainable and regenerative agriculture is that it provides us with new analytical frameworks for crop systems.
Again, crop mixtures provide an excellent illustrative example here.
Many researchers and farmers have been interested in growing crop mixtures, and many trials of crop mixtures have been undertaken, as summarised in a number of reviews and meta-analyses (e.g. Anil et al., 1998;Brooker et al., 2015;Kiaer et al., 2009;Li et al., 2020;Martin-Guay et al., 2018). However, when assessed across multiple trials although crop mixtures typically show positive effects on outcomes such as land use efficiency and weed, pest and disease suppression, the outcome of individual trials is often quite variable.
It proves hard then to come up with generic recommendations for farmers wanting to know which crop mixtures might work best in their system, or how to manage the crop to enhance the facilitative effects found in crop mixtures. Fortunately, ecological theory concerning facilitation provides us with analytical frameworks that can be applied to these challenges. An obvious example is the Stress Gradient Hypothesis (SGH; see Brooker et al., 2008;He et al., 2013 for overviews); put briefly, this states that the frequency or role of facilitative interactions increases in more severe (i.e. stressed or disturbed) environments. If facilitative interactions underpin many of the benefits of crop mixtures, the analytical framework provided by the SGH could be applied to analysing the outcome of crop mixture trials (e.g. Betencourt et al., 2012;Darch et al., 2018;Stefan et al., Submitted). In Figure 1, we illustrate this by taking the original U-shaped response curve of Bertness and Callaway (1994;Figure 1a) and replace the X-axis drivers with changes in agricultural systems or practices that could result from a shift towards more sustainable farming (Figure 1b). These include more stressful conditions for the crop, either directly because of reduced inputs, or because of the indirect consequences of this for negative biotic interactions with weeds, pests and diseases. In both these cases, the stress gradient hypothesis predicts an increased role of facilitative interactions.
Alternatively, rather than changes in agricultural practice, the X-axis could still be taken to represent environmental gradients in space or time (as in the original 1994 model), but with this variation occurring F I G U R E 1 Integration of sustainable crop production into the Stress Gradient Hypothesis (SGH). (a) The basic SGH concept, redrawn (and slightly adapted) from Bertness and Callaway (1994) and illustrating how the balance between the frequency of positive or negative interactions is expected to vary along gradients of physical stress or consumer pressure. The major types of facilitation relevant to these changes are indicated on the figure, and illustrated by a Scots pine Pinus sylvestris sapling being given protection from deer browsing by neighbouring heather, and a Silene acaulis cushion plant growing in the Dolomites, with other species growing within it. (b) Adaptation of the SGH model to focus on the key impacts in crop production systems of a shift towards more sustainable farming. In this case, increasing consumer (herbivore) pressure is replaced with increasing biotic stress from pathogens, pests and weeds as a consequence of reduced herbicide and pesticide use. Increasing physical stress results from reduced inputs of fertilisers and water. Indicated in the figure are examples of the major types of facilitative mechanisms which might become more frequent and important as we move along these abiotic or biotic stress gradients, illustrated by pea aphids and a barleypea intercrop. For more detail on these mechanisms, see text and Table 1 (a) (b) , e.g.
in crop systems as a result of climate change, land degradation or increased use of marginal land. To this end, we might predict then that the benefits of intercrops would be most noticeable under more stressful environmental conditions because of the enhanced role of facilitation.
We must point out that the SGH has provoked considerable debate, especially with respect to its applicability in productive environments (Maestre et al., 2009). Although a very substantial meta-analysis (He et al., 2013) has supported the SGH's proposed generic trends, there is still a need to test explicitly its applicability to the role of facilitation within multi-cultivar or multi-species crop systems.
At the same time, we need to address the issue of how to quantify facilitation between crop species in agricultural systems.  (Brown et al., 2017;Neugebauer et al., 2018). If we can identify species or cultivar mixtures that undertake particularly strong facilitative interactions, noting of course the challenges in assessing facilitation as outlined above, then we can test whether closely related cultivars or species also deliver such strong effects.

| SYNTHE S IS
Modern agriculture is facing many challenges. At least some farmers and some country's agricultural systems will likely become increasingly dependent on ecological processes and nature-based solutions as they try to move to a state of greater sustainability. This oversight is important. Knowing that facilitation plays a substantial role in the BEF relationships that help deliver sustainable agriculture provides us, first, with important new targets and research goals in terms of selecting traits and breeding for sustainability.
Second, it enables us to integrate fundamental ecological concepts with those from crop science, providing potentially important analytical frameworks that help us to better implement sustainable agricultural practice on the ground.
Moving forward, continued and enhanced collaborative working between ecologists and crop scientists will be essential, which, in turn, would enhance dialogue and mutual recognition of the understanding and approaches that both disciplines can bring to the challenge. This enhanced collaboration will enable us to harness the benefits of ecological knowledge for developing more sustainable agricultural practice. In particular, we need to move away from relying on generic increases in biodiversity to deliver facilitation-driven BEF effects. Instead, we need to continue efforts to understand the mechanistic basis of strong facilitative interactions in crop systems, and use this information to select and breed for improved combinations of genotypes and species, an approach which, in contrast to the purely genetics-based Genomic Selection of classic plant breeding, we have termed Functional Ecological Selection.

ACK N OWLED G EM ENTS
We would like to thank our Associate Editor and two reviewers for the very helpful and insightful comments on an earlier version of this paper. The contribution of James Hutton Institute staff was sup-

PEER R E V I E W
The peer review history for this article is available at https://publo ns.

DATA AVA I L A B I L I T Y S TAT E M E N T
This study does not include or use primary data.